Accept-Encoding parsing

[Mark Nottingham]

http.c parses the Accept-Encoding response header like this:

"""
if (!av_strcasecmp(tag, "Accept-Ranges") &&

!strncmp(p, "bytes", 5) &&
s->seekable == -1) {

h->is_streamed = 0;

} else if (!av_strcasecmp(tag, "Transfer-Encoding") &&

"""

This means that if the string "bytes" shows up in the header at all, it will match -- making it error-prone if a range using that sequence of characters is ever defined (e.g., "newbytes" as recently proposed on the HTTP WG mailing list).

The syntax isn't difficult to parse, it's a comma-separated list of tokens with optional whitespace (the same as several other headers):

​http://httpwg.github.io/specs/rfc7233.html#header.accept-ranges

(found by Rodger Combs on the IETF HTTP WG mailing list)

[Ridley Combs]

Correction: the existing implementation matches if the first 5 characters are "bytes", which means it wouldn't be compatible with a server that lists multiple allowed range units without starting with "bytes".